Boolean learning under noise-perturbations in hardware neural networks

Author:

Andreoli Louis1,Porte Xavier1,Chrétien Stéphane12,Jacquot Maxime1,Larger Laurent1,Brunner Daniel1

Affiliation:

1. FEMTO-ST/Optics Dept., UMR CNRS 6174, Univ. Bourgogne Franche-Comté, 15B avenue des Montboucons, 25030, Besançon Cedex, France

2. Laboratoire ERIC, UFR ASSP, Universite Lyon 2, 5 avenue Mendes France, 69676 Bron Cedex, France National Physical Laboratory, Teddington, UK The Alan Turing Institute, London, UK

Abstract

AbstractA high efficiency hardware integration of neural networks benefits from realizing nonlinearity, network connectivity and learning fully in a physical substrate. Multiple systems have recently implemented some or all of these operations, yet the focus was placed on addressing technological challenges. Fundamental questions regarding learning in hardware neural networks remain largely unexplored. Noise in particular is unavoidable in such architectures, and here we experimentally and theoretically investigate its interaction with a learning algorithm using an opto-electronic recurrent neural network. We find that noise strongly modifies the system’s path during convergence, and surprisingly fully decorrelates the final readout weight matrices. This highlights the importance of understanding architecture, noise and learning algorithm as interacting players, and therefore identifies the need for mathematical tools for noisy, analogue system optimization.

Funder

Region Bourgogne Franche-Comté

H2020 Marie Skłodowska-Curie Actions

Volkswagen Foundation

Publisher

Walter de Gruyter GmbH

Subject

Electrical and Electronic Engineering,Atomic and Molecular Physics, and Optics,Electronic, Optical and Magnetic Materials,Biotechnology

Reference60 articles.

1. Optical information processing based on an associative-memory model of neural nets with thresholding and feedback;Opt. Lett.,1985

2. Phase noise robustness of a coherent spatially parallel optical reservoir;IEEE J. Select.Top. Quant. Electron.,2020

3. Information processing using a single dynamical node as complex system;Nat. Commun.,2011

4. Reinforcement learning in a large scale photonic recurrent neural network;Optica,2018

5. Deep learning;Nature,2015

Cited by 12 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3